deepmd-python-inference

Name: deepmd-python-inference
Author: jinzhezenggroup

113stars23forks

How to install deepmd-python-inference?

jinzhezenggroup/computational-chemistry-agent-skills/deepmd-python-inference

$npx -y skills add jinzhezenggroup/computational-chemistry-agent-skills --skill deepmd-python-inference

Installs into the current project.

Summary

Run Python inference with DeePMD-kit models using the DeepPot API. Use when the user wants to load a trained/frozen DeePMD model (.pth or .pb) or a built-in pretrained model (e.g., DPA-3.2-5M) in Python, predict energy/force/virial for atomic configurations, evaluate descriptors,

Files · 1

View on GitHub

SKILL.md

1	# DeePMD-kit Python Inference
2
3	Load a trained DeePMD-kit model in Python and predict energy, forces, and virial for atomic configurations. Also covers CLI-based testing with `dp test`.
4
5	## Quick Start
6
7	```python
8	from deepmd.infer import DeepPot
9	import numpy as np
10
11	dp = DeepPot("model.pth")
12	coord = np.array([[1, 0, 0], [0, 0, 1.5], [1, 0, 3]]).reshape([1, -1])
13	cell = np.diag(10 * np.ones(3)).reshape([1, -1])
14	atype = [1, 0, 1]
15	e, f, v = dp.eval(coord, cell, atype)
16	```
17
18	## Agent Responsibilities
19
20	1. Determine the model source:
21	- Frozen model file (`.pth` for PyTorch, `.pb` for TensorFlow)
22	- Built-in pretrained model name (e.g., `DPA-3.2-5M`)
23	- Checkpoint file (requires freezing first)
24	1. Determine the inference task:
25	- Single-frame prediction (energy, force, virial)
26	- Batch prediction over multiple frames
27	- Descriptor evaluation
28	- Model deviation calculation
29	- CLI-based testing against labeled data
30	1. Help the user prepare input arrays in the correct format.
31	1. Run inference and report results.
32
33	## Python API: DeepPot
34
35	### Load a Model
36
37	```python
38	from deepmd.infer import DeepPot
39
40	# From a frozen PyTorch model
41	dp = DeepPot("model.pth")
42
43	# From a frozen TensorFlow model
44	dp = DeepPot("graph.pb")
45
46	# From a built-in pretrained model (auto-downloads if not cached)
47	dp = DeepPot("DPA-3.2-5M")
48	```
49
50	Built-in pretrained model names include `DPA-3.3-1M`, `DPA-3.2-5M`, `DPA-3.1-3M`, `DPA3-Omol-Large`, etc. DeePMD-kit will automatically download and cache the model on first use.
51
52	### Predict Energy, Forces, and Virial
53
54	```python
55	import numpy as np
56	from deepmd.infer import DeepPot
57
58	dp = DeepPot("model.pth")
59
60	# Prepare inputs
61	# coord: (nframes, natoms * 3) in Angstrom
62	# cell: (nframes, 9) cell vectors in Angstrom, row-major
63	# atype: list of atom type indices (length natoms)
64
65	coord = np.array(
66	[
67	[
68	0.0,
69	0.0,
70	0.0, # atom 0 (O)
71	0.0,
72	0.0,
73	1.0, # atom 1 (H)
74	0.0,
75	1.0,
76	0.0,
77	] # atom 2 (H)
78	]
79	).reshape([1, -1])
80
81	cell = np.diag([10.0, 10.0, 10.0]).reshape([1, -1])
82
83	# atype indices correspond to type_map order in the model
84	# e.g., if type_map = ["O", "H"], then O=0, H=1
85	atype = [0, 1, 1]
86
87	e, f, v = dp.eval(coord, cell, atype)
88
89	print(f"Energy (eV): {e}") # shape: (nframes, 1)
90	print(f"Forces (eV/A): {f}") # shape: (nframes, natoms, 3)
91	print(f"Virial (eV): {v}") # shape: (nframes, 9)
92	```
93
94	### Non-periodic Systems
95
96	For non-periodic (isolated) systems, pass `cell=None`:
97
98	```python
99	e, f, v = dp.eval(coord, None, atype)
100	```
101
102	### Batch Prediction
103
104	Process multiple frames at once:
105
106	```python
107	nframes = 10
108	natoms = 3
109
110	coords = np.random.rand(nframes, natoms * 3)
111	cells = np.tile(np.diag([10.0, 10.0, 10.0]).reshape([1, -1]), (nframes, 1))
112	atype = [0, 1, 1]
113
114	e, f, v = dp.eval(coords, cells, atype)
115	# e: (nframes, 1)
116	# f: (nframes, natoms, 3)
117	# v: (nframes, 9)
118	```
119
120	### Evaluate Descriptors
121
122	Extract the descriptor (atomic environment representation) from the model:
123
124	```python
125	descriptors = dp.eval_descriptor(coord, cell, atype)
126	# shape: (nframes, natoms, ndesc)
127	```
128
129	This can also be done via CLI:
130
131	```bash
132	dp eval-desc -m model.pth -s /path/to/system -o desc_output
133	```
134
135	### Calculate Model Deviation
136
137	Compare predictions from multiple models to estimate uncertainty:
138
139	```python
140	from deepmd.infer import calc_model_devi, DeepPot
141
142	coord = np.array([[1, 0, 0], [0, 0, 1.5], [1, 0, 3]]).reshape([1, -1])
143	cell = np.diag(10 * np.ones(3)).reshape([1, -1])
144	atype = [1, 0, 1]
145
146	graphs = [DeepPot("model_0.pth"), DeepPot("model_1.pth")]
147	model_devi = calc_model_devi(coord, cell, atype, graphs)
148	```
149
150	Important: avoid loading the same model multiple times in a loop, as this can cause memory leaks.
151
152	## CLI Testing: dp test
153
154	Test a frozen model against labeled data:
155
156	```bash
157	# Basic test
158	dp --pt test -m model.pth -s /path/to/test_system -n 30
159
160	# Test with detailed output
161	dp --pt test -m model.pth -s /path/to/test_system -n 30 -d test_detail
162	```
163
164	### dp test Options
165
166	\| Option \| Description \|
167	\| ---------------- \| ---------------------------------- \|
168	\| `-m MODEL` \| Path to the frozen model file \|
169	\| `-s SYSTEM` \| Path to the test data system \|
170	\| `-n NUMB` \| Number o

People who liked this also viewed

6 picks

Type

deepmd-python-inference

113stars23forks

View on GitHub

How to install deepmd-python-inference?

jinzhezenggroup/computational-chemistry-agent-skills/deepmd-python-inference

$npx -y skills add jinzhezenggroup/computational-chemistry-agent-skills --skill deepmd-python-inference

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# DeePMD-kit Python Inference
2
3	Load a trained DeePMD-kit model in Python and predict energy, forces, and virial for atomic configurations. Also covers CLI-based testing with `dp test`.
4
5	## Quick Start
6
7	```python
8	from deepmd.infer import DeepPot
9	import numpy as np
10
11	dp = DeepPot("model.pth")
12	coord = np.array([[1, 0, 0], [0, 0, 1.5], [1, 0, 3]]).reshape([1, -1])
13	cell = np.diag(10 * np.ones(3)).reshape([1, -1])
14	atype = [1, 0, 1]
15	e, f, v = dp.eval(coord, cell, atype)
16	```
17
18	## Agent Responsibilities
19
20	1. Determine the model source:
21	- Frozen model file (`.pth` for PyTorch, `.pb` for TensorFlow)
22	- Built-in pretrained model name (e.g., `DPA-3.2-5M`)
23	- Checkpoint file (requires freezing first)
24	1. Determine the inference task:
25	- Single-frame prediction (energy, force, virial)
26	- Batch prediction over multiple frames
27	- Descriptor evaluation
28	- Model deviation calculation
29	- CLI-based testing against labeled data
30	1. Help the user prepare input arrays in the correct format.
31	1. Run inference and report results.
32
33	## Python API: DeepPot
34
35	### Load a Model
36
37	```python
38	from deepmd.infer import DeepPot
39
40	# From a frozen PyTorch model
41	dp = DeepPot("model.pth")
42
43	# From a frozen TensorFlow model
44	dp = DeepPot("graph.pb")
45
46	# From a built-in pretrained model (auto-downloads if not cached)
47	dp = DeepPot("DPA-3.2-5M")
48	```
49
50	Built-in pretrained model names include `DPA-3.3-1M`, `DPA-3.2-5M`, `DPA-3.1-3M`, `DPA3-Omol-Large`, etc. DeePMD-kit will automatically download and cache the model on first use.
51
52	### Predict Energy, Forces, and Virial
53
54	```python
55	import numpy as np
56	from deepmd.infer import DeepPot
57
58	dp = DeepPot("model.pth")
59
60	# Prepare inputs
61	# coord: (nframes, natoms * 3) in Angstrom
62	# cell: (nframes, 9) cell vectors in Angstrom, row-major
63	# atype: list of atom type indices (length natoms)
64
65	coord = np.array(
66	[
67	[
68	0.0,
69	0.0,
70	0.0, # atom 0 (O)
71	0.0,
72	0.0,
73	1.0, # atom 1 (H)
74	0.0,
75	1.0,
76	0.0,
77	] # atom 2 (H)
78	]
79	).reshape([1, -1])
80
81	cell = np.diag([10.0, 10.0, 10.0]).reshape([1, -1])
82
83	# atype indices correspond to type_map order in the model
84	# e.g., if type_map = ["O", "H"], then O=0, H=1
85	atype = [0, 1, 1]
86
87	e, f, v = dp.eval(coord, cell, atype)
88
89	print(f"Energy (eV): {e}") # shape: (nframes, 1)
90	print(f"Forces (eV/A): {f}") # shape: (nframes, natoms, 3)
91	print(f"Virial (eV): {v}") # shape: (nframes, 9)
92	```
93
94	### Non-periodic Systems
95
96	For non-periodic (isolated) systems, pass `cell=None`:
97
98	```python
99	e, f, v = dp.eval(coord, None, atype)
100	```
101
102	### Batch Prediction
103
104	Process multiple frames at once:
105
106	```python
107	nframes = 10
108	natoms = 3
109
110	coords = np.random.rand(nframes, natoms * 3)
111	cells = np.tile(np.diag([10.0, 10.0, 10.0]).reshape([1, -1]), (nframes, 1))
112	atype = [0, 1, 1]
113
114	e, f, v = dp.eval(coords, cells, atype)
115	# e: (nframes, 1)
116	# f: (nframes, natoms, 3)
117	# v: (nframes, 9)
118	```
119
120	### Evaluate Descriptors
121
122	Extract the descriptor (atomic environment representation) from the model:
123
124	```python
125	descriptors = dp.eval_descriptor(coord, cell, atype)
126	# shape: (nframes, natoms, ndesc)
127	```
128
129	This can also be done via CLI:
130
131	```bash
132	dp eval-desc -m model.pth -s /path/to/system -o desc_output
133	```
134
135	### Calculate Model Deviation
136
137	Compare predictions from multiple models to estimate uncertainty:
138
139	```python
140	from deepmd.infer import calc_model_devi, DeepPot
141
142	coord = np.array([[1, 0, 0], [0, 0, 1.5], [1, 0, 3]]).reshape([1, -1])
143	cell = np.diag(10 * np.ones(3)).reshape([1, -1])
144	atype = [1, 0, 1]
145
146	graphs = [DeepPot("model_0.pth"), DeepPot("model_1.pth")]
147	model_devi = calc_model_devi(coord, cell, atype, graphs)
148	```
149
150	Important: avoid loading the same model multiple times in a loop, as this can cause memory leaks.
151
152	## CLI Testing: dp test
153
154	Test a frozen model against labeled data:
155
156	```bash
157	# Basic test
158	dp --pt test -m model.pth -s /path/to/test_system -n 30
159
160	# Test with detailed output
161	dp --pt test -m model.pth -s /path/to/test_system -n 30 -d test_detail
162	```
163
164	### dp test Options
165
166	\| Option \| Description \|
167	\| ---------------- \| ---------------------------------- \|
168	\| `-m MODEL` \| Path to the frozen model file \|
169	\| `-s SYSTEM` \| Path to the test data system \|
170	\| `-n NUMB` \| Number o

People who liked this also viewed

6 picks

Type